Metaculus

Metaculus is a forecasting platform founded in 2015 that aggregates user predictions to generate community probability estimates on a wide range of questions, from scientific milestones to geopolitical events to AI capabilities. Unlike financial prediction markets, Metaculus is non-monetary — participants earn reputation points and forecasting scores rather than cash, though this has not diminished the quality of its predictions.

Metaculus questions tend to be longer-horizon and more nuanced than financial prediction markets. Where Polymarket might focus on "Will X candidate win the next election?", Metaculus hosts questions like "Will GPT-5 score above 90th percentile on the bar exam by 2025?" or "When will nuclear fusion produce net energy commercially?" The platform has published impressive track records in domains like COVID-19 modeling and AI development timelines.

The platform uses a Bayesian aggregation algorithm to weight forecasters' contributions by their historical accuracy, meaning the community prediction is not a simple average but a skill-weighted consensus. This approach has shown strong calibration in retrospective analyses and has earned Metaculus citations in academic research on forecasting methodology.